EP2200163A1 - Method for predictive PWM current control - Google Patents

Abstract

A new and advanced PWM current control is provided by a new method and control system architecture that achieve the very high performances of an advanced current control for full-bridge stages, in terms of accuracy, error, speed and frequency response, but with an outstandingly reduced complexity in terms of analog circuits required, comparable with that of an elementary peak current control. The only analog blocks required are one current sense transducer (i.e. a series resistor or a sense-FET) and one comparator for the current sensing while the rest of the control circuitry is digital.

Description

• The present disclosure relates in general to switching mode driving of electrical loads and in particular of inductive loads, through an output power stage. The invention provides an efficient and effective method of controlling the load current in presence of a significant and inconstant ripple.
BACKGROUND
• In common current control systems using the technique of peak control, a current sense transducer is used in order to sense the current in the output power bridge and an analog comparator compares the sense voltage with a reference voltage related to the desired current for the load. This elementary system is just able to assess whether the current is above or below a certain threshold. In a power bridge using as sensing element a series resistor connected between the low side switches and ground, the current information is not always available, in fact it depends on the conditions of current recirculation.
• Typically, a peak current control system let the load current increase until it reaches a set threshold and then turns off the power stage and the current decays for a time that depends on the type of control that is implemented: Toff-constant or constant PWM frequency, as typical examples. In FIG. 1 a Toff-constant mode is shown.
• The advantage of this solution is the evident simplicity of the implementing circuits. A disadvantage is that the system is affected by systematic error caused by changes of the current ripple that notably increases with the absolute value of the current and is proportionally tied to the PWM period.
• The publication "Predictive Digital Current Programmed Control", Jingquan Chen, et al., Jan. 2003, Power Electronics, IEEE Transactions, Vol. 18, discloses a predictive digital current control for "valley", "peak" or "average current" in basic converters topologies: buck, boost and buck-boost. The work reveals that for each parameter of interest there is a choice of most appropriate pulse-width modulation for accomplishing a predictive digital current control that avoids oscillation problems. The optimal duty cycle is predictively computed by the controller, on the basis of a number of electrical parameters such as: load current, input voltage, output voltage and inductance value, implying multiplications and divisions.
SUMMARY OF THE INVENTION
• The novel current control method and architecture being disclosed maintain the basic principle of a current peak control, thus requiring an analog circuital part composed of just a sense element and a sense comparator, but effects a predictive control that is carried out with simple time measurements and algebraical additions, implemented with relatively simple digital circuitry.
• Basically, the method of this disclosure comprises the steps of measuring the time interval from the initial instant of an active ON-state of the power stage of an ON-phase to the instant the load current reaches the reference current value as determined by a valid switching of the sense comparator, so intended when the sensed current is below the reference threshold at the beginning of the ON phase, and, differently from the known techniques, instead of switching off the power stage continuing the ON-phase by maintaining in a predictive ON-state the power stage, for a further time interval equal to an average between the just measured active ON-state interval and the active ON-state interval measured during a past PWM cycle that may be the cycle immediately preceding the current cycle or a more remote past cycle, for example distant 3 to 10 or more cycles from the current cycle, or even an ON-state interval value that is averaged over a number of preceding cycles. The choice of the second term of the average may be fixed or programmable, depending on characteristics of the application, and even adaptively set in function of load conditions.
• Thereafter, the power stage is switched off for a fixed interval time, as done in a Toff constant control.
• During such a predictive ON-state, a sort of "blind" driving is performed because the load current has surpassed the reference threshold and the comparator has already triggered. During this protracted drive interval the comparator output is ignored as during the following OFF-phase of the PWM cycle.
• The control system of this invention produces a fast convergence of the average load current towards the set reference threshold value that triggers the comparator dramatically improving the accuracy of the control system, compared with a common current peak control system. In fact, the average value of a triangular wave as the ripple, coincides with the mid points of each of its segments and, at steady state, the system tends to equalize the duration of the active Ton time and the predictive Ton_pred time.
• Initialization, implementation of a blanking time to prevent spurious switching and a current check at turn-on further enhance performance.
• Moreover, procedures of simple implementation for changing the set reference threshold and for achieving a quasi-constant PWM frequency may expand flexibility and utility of the novel control system for a broad range of applications.
• According to a further embodiment, an adaptive current decay mode selection during the Toff phase of the PWM cycle may be implemented in an outstandingly simplified manner for further improving performances.
BRIEF DESCRIPTION OF THE DRAWINGS
• FIG. 1 illustrates a classic peak current control.
• FIG. 2 illustrates a basic predictive current control of the present invention.
• FIG. 3 illustrates a phase of convergence of the current control toward steady state condition.
• FIG. 4 shows a circuit diagram and related signal waveforms of an embodiment of the predictive current control of this invention.
• FIG. 5 is a simplified basic flow chart of a basic embodiment of the predictive current control.
• FIG. 6 is a flow chart for a preferred embodiment of the predictive current control.
• FIG. 7 is a diagram illustrating the initialization of the control system at turn-on of the output power stage.
• FIG. 8 is a diagram illustrating the introduction of a blanking time interval and of a current check at turn-on of the power stage.
• FIG. 9 is a diagram illustrating an event of change of a set current reference value (incrementing adjustment).
• FIG. 10 is a diagram illustrating an event of change of a set current reference value (decrementing adjustment).
• FIG. 11 illustrates schemes of slow decay mode and fast decay mode of the load current during an off-phase of PWM driving.
• FIG. 12 is a diagram of the functioning of a sub-system of adaptive decay mode selection.
• FIG. 13 is the flow chart of the sub-system of adaptive decay mode selection.
• FIG. 14 illustrates an example of adaptive current decay mode selection in a motion control application.
DETAILED DESCRIPTION OF EMBODIMENTS
• FIG. 2 diagrammatically specifies the temporal quantities that are utilized by the basic algorithm of the current control method of this invention.
• The current control method of this disclosure exploits the calculation power of digital electronics for real time measuring of the time interval Ton(n) from the initial instant of an active ON-state of the power stage of an ON-phase, to the instant the load current reaches the reference current value as determined by a valid switching of the sense comparator. Differently from a classical peak current control method, instead of switching off the power stage when the sense comparator triggers, the power stage is left continuing the ON-phase, by maintaining it in a "predictive ON-state" for a further time interval Tpred(n) equal to the average between the just measured active ON-state interval Ton(n) and the active ON-state interval Ton(n-1) measured during the precedent PWM cycle. For sake of simplicity, in the following analysis and in the figures, it is assumed that the past PWM cycle be the last one (n-1) before the current cycle (n), however as already said the second term for the averaging may be any other appropriate value related to a more remote past PWM cycle or even a "short-range" average value over several past PWM cycles.
• Thereafter, the power stage is switched off for a fixed interval time Toff, as done in a constant Toff control mode.
• During such a predictive ON-state, a sort of "blind" driving is performed because the load current has surpassed the reference threshold and the comparator has already triggered. During this protracted drive interval the comparator output state is ignored as during the following OFF-phase of the PWM cycle.
• The predictive algorithm of the current control method of this invention provides a fast and safe convergence to the reference value. In mathematical terms it can be demonstrated as showed below.
• For the system to converge in steady state as depicted in FIG. 3 , it must be: $$\mathrm{\Delta }{I}^{+}=\mathrm{\Delta }{I}^{-}$$

  

  
  where: $$\begin{array}{c}\mathrm{\Delta }{I}^{+}=\left(\frac{\mathit{d}{\mathit{I}}_{\mathit{ON}}}{\mathit{dt}}\right)\cdot \left(T_{\mathit{ON}}+T_{\mathit{PRED}}\right)=K_{\mathit{ON}}\cdot \left(T_{\mathit{ON}}+T_{\mathit{PRED}}\right)\\ \mathrm{\Delta }{I}^{-}=\left(\frac{\mathit{d}{\mathit{I}}_{\mathit{OFF}}}{\mathit{dt}}\right)\cdot T_{\mathit{OFF}}=K_{\mathit{OFF}}\cdot T_{\mathit{OFF}}\end{array}$$

  
• In PWM control, the dI/dt can be considered substantially constant during a single PWM cycle, Therefore: $$\begin{array}{c}\mathrm{\Delta }{I}^{+}=K_{\mathit{ON}}\cdot \left(T_{\mathit{ON}}+T_{\mathit{PRED}}\right)\\ \mathrm{\Delta }{I}^{-}=K_{\mathit{OFF}}\cdot T_{\mathit{OFF}}\end{array}$$

  

  
  and: $$\begin{array}{c}\mathrm{\Delta }{I}_{\mathit{ON}}=K_{\mathit{ON}}\cdot T_{\mathit{ON}}\\ \mathrm{\Delta }{I}_{\mathit{PRED}}=K_{\mathit{ON}}\cdot T_{\mathit{PRED}}\\ \mathrm{\Delta }{I}_{\mathit{OFF}}=K_{\mathit{OFF}}\cdot T_{\mathit{OFF}}\end{array}$$

  
• As observed in FIG. 3 , the relationship between the current variation during the ON time at a certain point and the variation during the previous predictive and OFF times is the following: $$\mathrm{\Delta }{I}_{\mathit{ON}}\left(n+1\right)=\mathrm{\Delta }{I}_{\mathit{OFF}}-\mathrm{\Delta }{I}_{\mathit{PRED}}\left(n\right)=\mathrm{\Delta }{I}_{\mathit{OFF}}-\frac{K_{\mathit{ON}}}{2}\cdot \left(T_{\mathit{ON}}\left(n\right)+T_{\mathit{ON}}\left(n-1\right)\right)$$

  
• This can be rewritten as the following numeric recursive succession describing how the ON time varies at each PWM cycle: $$\begin{array}{|c|}\hline T_{\mathit{ON}}\left(n+1\right)=\frac{{\mathit{K}}_{\mathit{OFF}}}{{\mathit{K}}_{\mathit{ON}}}\mathit{\cdot }{\mathit{T}}_{\mathit{OFF}}-\frac{T_{\mathit{ON}}\left(n\right)+T_{\mathit{ON}}\left(n-1\right)}{2}\\ \hline\end{array}$$

  
• Where
  $\frac{K_{\mathit{OFF}}}{K_{\mathit{ON}}}\mathrm{.}$

  

  T_OFF is a constant that depends from the dI/dt and the OFF time set by the control system.
• This is a typical recursive numeric succession (c being a constant) that has the following convergence value: $$\begin{array}{|c|}\hline a\left(n+1\right)=c-a\left(n\right)-a\left(n-1\right)\to \underset{n\to \infty }{\lim }a\left(n\right)=\frac{\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="imgf0002.png,imgf0003.png"\; state="\{\{state\}\}">c</figure-callout>}}{2}\\ \hline\end{array}$$

  
• Therefore, the system converges to the final value: $$\begin{array}{|c|}\hline T_{\mathit{ON}}\left(n\right)={\mathit{T}}_{\mathit{ON}}\left(n-1\right)=\frac{1}{2}\cdot \left(\frac{K_{\mathit{OFF}}}{K_{\mathit{ON}}}\right)\cdot T_{\mathit{OFF}}\\ \hline\end{array}$$

  
• In the embodiment described above, the current control system is of the type denominated Toff-constant type PWM control, though different types of control may be alternatively implemented, as will be illustrated in the ensuing description.
• FIG. 4 shows an exemplary circuit embodying the current control system of this invention and the relevant signal waveforms of the state machine represented by the block DIGITAL CONTROL of the circuit diagram.
• A full-bridge controls in both directions the current in an inductive load. The analog part ANALOG BLOCK of the control system, commonly include a SENSE ELEMENT that may even be a simple resistor or a sense FET, and a comparator ANALOG COMPARATOR, OR any equivalent device able to compare the current in the load with a given reference.
• A first remark is that the novel current control system of this disclosure, according to a basic embodiment, doesn't require to sense the current in the load when the full-bridge is in OFF-state, because the controller DIGITAL CONTROL works only on feedback information produced during the ON-state of the output power stage.
• According to the basic embodiment, the sense resistor would not used to produce information on the load current when the current is decaying (e.g. eventually in slow or fast mode).
• Therefore, the analog comparator merely provides a comparison signal of the current delivered to the load with a given reference threshold. The output of the comparator is fed the control logic of the system, DIGITAL CONTROL, which outputs the driving signals to the power switches of the full-bridge.
FIG. 5 is a flow chart of the basic predictive current control method:
• The flow chart describes the principle of working of the state machine which is the core of the predictive current control method. According to the circuit diagram and related waveforms of FIG. 4, the flow chart illustrates that the only input variable to the state machine is the output of the sense comparator, which is only able to signal whether the current is above or below the reference value, when the power bridge is in ON-state. Both the phases of predictive ON-state and of OFF-state are completely blind as far as input of the control system is concerned.
• According to a preferred embodiment, a more advanced control may be implemented in order to prevent possible perturbations and/or improve performance when far from steady state or on account of non-idealities of the electrical components used.
• A flow chart of an exemplary embodiment of a more advanced control is shown in FIG. 6 .
Initialization
• The flow chart of FIG. 6, contemplates the initialization with a default value of the Ton register registers that will thereafter be updated to contain the ON-state time measurement of the last PWM cycle, and at start up is used as predictive Ton time, in order to speed up the convergence process, as diagrammatically illustrated in FIG. 7 .
Blanking time mask
• At every power bridge switching to an ON-state (turn-ON), a blanking time mask should be introduced in order to avoid spurious triggering of the sense comparator because of spikes or perturbations caused power switch transitions. After this blanking time is elapsed, the power switches can be considered completely turned ON and thence the output of the comparator is considered valid. If a relatively long blanking time is implemented for robustness purposes, it is possible to consider this blanking time as the initial value of the Ton time counter, for compensating any error in the ON-time measurement and thence in converging to the reference current value.
Current check at turn-ONs
• Upon elapsing of a blanking time, the output COMPOUT of the sense comparator is made visible to the digital control circuitry that checks if the current is above or below the reference threshold. If the current is found above the threshold it means that the current has increased too much during the blanked ON-state and this could happen because of different reasons. For example, during the preceding OFF-phase the current has decayed insufficiently or the current reference has been reduced.
• In this case, immediately after the blanking time, the control system set the power stage in OFF-state in order to allow current to decrease further down of the reference value, skipping completely the ON-phase of the PWM cycle, that is the active ON-state and the ensuing predictive-ON state.
• By contrast, if after the blanking time the current is found to be below the reference threshold, the system follows the normal routine, setting the power bridge in ON-state, measuring the Ton interval and after it calculates Tpred, continuing to deliver current to the load for the calculated predictive ON-state Tpred interval. The PWM OFF-phase will follows the predictive ON-state, for a certain constant duration Toff, and after the control system will again turn the power bridge ON and wait for the elapsing of the constant blanking time. The Toff time and the blanking time are constant values set by the user and usually will depend from the application features.
• It should be remarked that the flow charts of FIG. 5 and FIG. 6 should be considered exemplary as far as basic illustrations of embodiments of the method of this invention, indeed the sequence of instructions and the relative positions of some of the controls that are implemented may be different, what really is important is the result of such state machine operations.
• Fig. 8 is a time diagram illustrating how a current control system of this invention provided with blanking time and turn-on current check functions rapidly converges to the reference current value (in the sense of rapidly reaching steady state).
Change of the reference value
• When the control logic changes the current reference value, the Ton time measurement performed during the ON-state is ignored until the next valid switching of the sense comparator and the predictive Ton time retains the last validly calculated value. This avoids Ton time measurements much longer than the real Ton steady state value, which may happen soon after the reference value change.
• Neglecting the Ton time measurement immediately after a variation of the current reference value, avoids perturbations in the predictive current control loop and reduces the time required to converge to the new desired steady state.
• As illustrated in the flow chart of FIG.6 and in the diagrams of FIG. 9, for the case of a decrementing adjustment of the reference value, and of FIG. 10, for the case of an incrementing adjustment of the reference value, just after a current reference change, the last value of the predictive Ton time is retained unchanged until the next valid sense comparator triggering.
• It may be observed that, in case of a decrementing change of the reference value, the sensed current may be above the new current reference, however, in this case, the blanking time routine sets the control system immediately back to OFF state for a further Toff time and the Ton time measurement is ignored.
• In any case, a sense comparator triggering is considered valid for starting the predictive ON-state only when the sensed current is below the reference threshold at the beginning of the ON phase.
Quasi constant PWM frequency
• As explained above, after few cycles an embodiment of the novel system of this invention reaches steady state condition and the resulting Ton will be given by the following equation: $$\begin{array}{|c|}\hline T_{\mathit{ON}}\left(n\right)={\mathit{T}}_{\mathit{ON}}\left(n-1\right)=\frac{1}{2}\cdot \left(\frac{K_{\mathit{OFF}}}{K_{\mathit{ON}}}\right)\cdot T_{\mathit{OFF}}\\ \hline\end{array},$$

  

  
  where Toff may be a constant time defined by the user.
• In order to converge properly Toff should remain constant in approaching the convergence value, as far as the reference value is fixed or practically so.
  However, when the reference value changes, it is possible to modify also Toff, in view of the fact that in any case the system would be perturbed.
• Because the predictive ON-state is based on the measurement of Ton during a past PWM cycle and the reference value adjustments are normally relatively small in absolute value, it is worthwhile to avoid large variations of the Toff parameter, considering that Toff is strictly related to the switching frequency of the PWM control, which, once fixed the current reference, depends only by Toff, as per the following equation: $$\begin{array}{|c|}\hline f_{\mathit{SW}}=\frac{1}{T_{\mathit{ON}}+T_{\mathit{PRED}}+T_{\mathit{OFF}}}=\frac{1}{2\cdot T_{\mathit{ON}}+T_{\mathit{OFF}}}=\frac{1}{\frac{K_{\mathit{OFF}}}{K_{\mathit{ON}}}\cdot T_{\mathit{OFF}}+T_{\mathit{OFF}}}=\frac{1}{\left(\frac{K_{\mathit{OFF}}}{K_{\mathit{ON}}}+1\right)\cdot T_{\mathit{OFF}}}=\frac{K_{\mathit{ON}}}{K_{\mathit{OFF}}+K_{\mathit{ON}}}\cdot \frac{1}{T_{\mathit{OFF}}}\\ \hline\end{array}$$

  
• Because after a small current reference variation the convergence value of Ton would be quite close to the previously measured Ton value, the time Toff could be advantageously selected (one time only after a reference value change) by the following approximately determined value: $$\begin{array}{|c|}\hline T_{\mathit{OFF}}=\frac{1}{f_{\mathit{SW}}}-{\mathit{T}}_{\mathit{ON}}\left(n\right)\\ \hline\end{array},$$

  

  
  where Ton(n) is the last valid measurement of the duration of the ON-state, with the previous current reference value.
• In this way, the real switching frequency will remain very close to the design value and the novel current control, differently from a classic peak current control, can be considered as a quasi-constant PWM frequency control system.
Adaptive current decay selection
• The slope of load current decay during OFF-phases of PWM driving may be selected from different decay modes, generally between a slow decay and a fast decay.
• In FIG. 11 two different decay modes are shown for the case of a full-bridge power stage. The same definitions apply to other power stage topologies, for example a 3-phase bridge and others.
• A slow decay is implemented when the load is short-circuited, voltage drop on the load is minimized and the current discharges on passive components. A fast decay is implemented when the current in the load flows through the switches towards the supply nodes and the voltage drop on the load approximates the supply voltage.
• The definition slow and fast decay refers to the different current slopes as shown in the following relationships: $$\frac{\mathrm{\Delta }{I}_{\mathit{SLOW}}}{\mathrm{\Delta }T}=\frac{{\mathit{V}}_{\mathit{DS\_switch}}\mathit{+}{\mathit{V}}_{\mathit{DIODE}}}{{\mathit{L}}_{\mathit{LOAD}}}$$

  

  $$\frac{\mathrm{\Delta }{I}_{\mathit{FAST}}}{\mathrm{\Delta }T}=\frac{{\mathit{V}}_{\mathit{BUS}}\mathit{+}2\mathit{\cdot }{\mathit{V}}_{\mathit{DIODE}}}{{\mathit{L}}_{\mathit{LOAD}}}$$

  
• In common approaches, the decay mode is fixed and can be selected by the external user/controller depending on the performance of the control system. Normally the slow decay is the most often used mode because it provides lower PWM frequencies (and therefore lower power dissipation at switching) in fixed off-time control systems, or lower current ripple in fixed PWM frequency control systems.
• But there are also other different solutions which make use of a mixed approach to the current decay, known as mixed decay. A known controller of this kind is the Allegro 3977 IC. It uses a mixed decay mode with fixed off-time in its PWM current regulators, which limits the peak current in the load to a reference threshold. Initially, a diagonal pair of the power bridge is enabled and current flows through the inductive load as shown in FIG. 11 (a). When the load current reaches the desired value, a current-sense comparator resets the PWM latch, which turns off both the high side and the low side switch in order to obtain mixed decay mode, and the current recirculates as shown in FIG. 11(b) and (c) . During this recirculation the current decreases until the fixed off time expires.
• Mixed decay splits the fixed off-time of the PWM cycle into fast and then slow decay. After the fixed off-time of the PWM cycle, the appropriate switches are enabled again, the inductive load current increases and the PWM cycle is repeated. Using mixed decay with fixed off time has the advantage that PWM frequency is variable (lower peak in EMC spectrum, because energy in spectrum is smeared), but it needs a relatively high frequency to guarantee operation above 20 kHz and this generates additional heat-losses.
• Other known controllers of this kind are deployed in Infineon TLE-472x series ICs, and in Toshiba TB62200. The latter uses fixed frequency PWM with slow, fast and mixed decay. In mixed decay mode, additional switching points are required which entails increased complexity.
• All drive systems working in slow decay mode have the problem of the minimum ON-phase time that can be handled. In fact in a typical power application using PWM current control, it is impossible to accomplish a zero percent duty cycle. The reason being power bridge delays that limit the minimum ON-phase time to a certain value. In some conditions the slow decay mode is unable to guarantee proper current regulation, because the OFF-phase time is no longer sufficient to decay the current down to a proper value and consequently the ON-phase time should become shorter that the viable minimum.
• In practice, a typical condition in which the system tends toward very low duty cycles is when the set reference current value is very low. The situation could be even more critical if the supply voltage of the power bridge is relatively low and the load inductance and resistance are high, which means that the current slopes are also significantly slow.
• Considering the power bridge of FIG. 11 , in first approximation, once a preferred reference current value is reached, the effective duty cycle applied to the load is: $$D\cong \frac{\overline{V_{\mathit{LOAD}}}}{V_{\mathit{SUPPLY}}}=\frac{\overline{I_{\mathit{LOAD}}}\cdot R_{\mathit{LOAD}}}{V_{\mathit{SUPPLY}}},$$

  
• The lower the average load current is, the lower will be the effective duty cycle and therefore the ON-phase duration.
• This holds whether the PWM control is a fixed OFF-time or a fixed frequency one.
• When the ON-time duration reaches the viable minimum value, in order to prevent impairing of performance, the system should switch to a fast decay mode for making faster the decay of the current and greater the duty cycle.
• A novel and outstandingly effective current decay architecture, thought to be eventually introduced in a novel current control system of this invention to further enhance its flexibility of application to a broad spectrum of systems, provides for an intelligent (adaptive) selection of the current decay mode, either slow or fast, depending on the current condition in the load for better adapting the driving to every condition of operation.
• Alike for the current control algorithm of this invention, also the ancillary adaptive current decay mode selection is based on the measurement of the ON-phase duration.
• Basically, the sub-system measures cycle by cycle the ON-phase duration and when the measured time becomes shorter than a programmed threshold, the system passes from a slow decay mode to a fast decay mode.
• The above mentioned programmed threshold may correspond to the blanking time used by the current control system for masking spurious signals caused by power stage transitions.
• FIG. 12 and FIG. 13 are an exemplary scheme and flow chart of an embodiment of an ancillary adaptive current decay mode selection sub-system.
• Apart from being implementable in a current control system of this invention with minimum added complexity, an advantage of this new decay mode selection architecture is its outstanding suitability for adaptive decay selection. The sub-system retains its high efficiency in most critical applications where the load may not be completely passive, as for example in motion control applications. In fact in these applications, the load cannot be simply equated to an inductor in series with a resistor but a voltage generator representing the BEMF source must be considered in series to the load. This is because the rotational energy stored in the motor works as a generator on the load and it is able to alter the current waveform during current decay, as illustrated in FIG. 14 .
• The decay mode adaptive selection sub-system can be applied to any kind of PWM current control such for example to a classical peak current control system using fixed OFF time technique or fixed PWM frequency technique.

Claims (9)

1. A method of predictive current control for driving in a PWM mode a load through a power stage, alternating ON-phases and OFF-phases of the power stage, and using a load current sense element and a sense comparator, comprising the steps of:

   setting a comparator threshold corresponding to a certain reference current value;

   measuring the time interval from the initial instant of an active ON-state of an ON-phase of the power stage to the instant the load current reaches said reference current value as determined by a switching of said sense comparator, and continuing the ON-phase by maintaining in a predictive ON-state the power stage for a further time interval equal to an average between the just measured active ON-state time interval and the active ON-state time interval measured during a past PWM cycle.
2. The method of claim 1, wherein after the predictive ON-state the power stage is set in OFF-phase for a constant Toff time.
3. The method of claim 1, wherein a default active ON-state time interval is used as said average time determination at power on.
4. The method of claim 1, wherein the active ON-state time interval measured during a past PWM cycle relates to either a past PWM cycle immediately preceding the current cycle or a more remote past cycle, or an ON-state interval time value averaged over a number of preceding PWM cycles.
5. The method of claim 1, wherein a blanking time is implemented at every turn-on of the power stage to avoid spurious triggering of the sense comparator due to spikes or perturbations caused by power stage transitions.
6. The method of claim 5, wherein at the end of said blanking time, the output of the sense comparator provides information on whether the current is above or below said threshold;
   if above, the ON-phase is aborted and the power stage is set in an OFF-state; if below, the ON-phase is performed followed by an OFF-phase before turning on again the power stage for a new ON-phase and waiting for the elapsing of said blanking time.
7. The method of claim 1, wherein when said reference current value is modified, any active ON-state time measurement is ignored until a next valid switching of the sense comparator and the predictive ON-state time interval retains the last validly calculated value.
8. The method of claim 1, further including adaptive current decay mode selection for OFF-phases comprising the steps of
   computing the duration of each ON-phase;
   comparing the duration with a programmed minimum threshold;
   either maintaining a default slow decay mode if the duration of the ON-phase is longer than said threshold or switching to a fast decay mode if the duration is shorter than said threshold.
9. A method of adaptive current decay mode selection in driving in a PWM mode a load through a power stage, alternating ON-phases and OFF-phases of the power stage, and using a load current sense element and a sense comparator for providing input information to a digital control system, comprising the steps of
   computing the duration of each ON-phase;
   comparing the duration with a programmed minimum threshold;
   either maintaining a default slow decay mode if the duration of the ON-phase is longer than said threshold or switching to a fast decay mode if the duration is shorter than said threshold.

Images